Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,114 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import gradio as gr
|
3 |
+
from PIL import Image
|
4 |
+
from rembg import remove
|
5 |
+
from diffusers import StableDiffusionPipeline
|
6 |
+
|
7 |
+
# -----------------------------------------------------------------------------
|
8 |
+
# Helper function to adjust image size to multiples of 8.
|
9 |
+
# -----------------------------------------------------------------------------
|
10 |
+
def adjust_size(w, h):
|
11 |
+
"""
|
12 |
+
Adjust width and height to be multiples of 8, as required by the Stable Diffusion model.
|
13 |
+
"""
|
14 |
+
new_w = (w // 8) * 8
|
15 |
+
new_h = (h // 8) * 8
|
16 |
+
return new_w, new_h
|
17 |
+
|
18 |
+
# -----------------------------------------------------------------------------
|
19 |
+
# Core processing function:
|
20 |
+
# 1. Remove background from the uploaded image.
|
21 |
+
# 2. Generate a new background image based on the text prompt.
|
22 |
+
# 3. Composite the foreground onto the generated background.
|
23 |
+
# -----------------------------------------------------------------------------
|
24 |
+
def process_image(input_image: Image.Image, bg_prompt: str) -> Image.Image:
|
25 |
+
"""
|
26 |
+
Processes the uploaded image by removing its background and replacing it with a generated one.
|
27 |
+
|
28 |
+
Parameters:
|
29 |
+
input_image (PIL.Image.Image): The uploaded image.
|
30 |
+
bg_prompt (str): Text prompt describing the new background.
|
31 |
+
|
32 |
+
Returns:
|
33 |
+
PIL.Image.Image: The final composited image.
|
34 |
+
"""
|
35 |
+
if input_image is None:
|
36 |
+
raise ValueError("No image provided.")
|
37 |
+
|
38 |
+
# Step 1: Remove the background from the input image.
|
39 |
+
print("Removing background from the uploaded image...")
|
40 |
+
foreground = remove(input_image)
|
41 |
+
foreground = foreground.convert("RGBA")
|
42 |
+
|
43 |
+
# Step 2: Determine new dimensions (multiples of 8) based on the foreground.
|
44 |
+
orig_w, orig_h = foreground.size
|
45 |
+
gen_w, gen_h = adjust_size(orig_w, orig_h)
|
46 |
+
print(f"Original size: {orig_w}x{orig_h} | Adjusted size: {gen_w}x{gen_h}")
|
47 |
+
|
48 |
+
# Step 3: Generate a new background using the provided text prompt.
|
49 |
+
print("Generating new background using Stable Diffusion...")
|
50 |
+
bg_output = pipe(
|
51 |
+
bg_prompt,
|
52 |
+
height=gen_h,
|
53 |
+
width=gen_w,
|
54 |
+
num_inference_steps=50, # Adjust as needed.
|
55 |
+
guidance_scale=7.5 # Adjust for prompt adherence.
|
56 |
+
)
|
57 |
+
# The generated background is in RGB mode; convert to RGBA for compositing.
|
58 |
+
background = bg_output.images[0].convert("RGBA")
|
59 |
+
|
60 |
+
# Step 4: If necessary, resize the foreground to match the background.
|
61 |
+
if foreground.size != background.size:
|
62 |
+
print("Resizing foreground to match background dimensions...")
|
63 |
+
foreground = foreground.resize(background.size, Image.ANTIALIAS)
|
64 |
+
|
65 |
+
# Step 5: Composite the foreground over the new background.
|
66 |
+
print("Compositing images...")
|
67 |
+
final_image = Image.alpha_composite(background, foreground)
|
68 |
+
|
69 |
+
return final_image
|
70 |
+
|
71 |
+
# -----------------------------------------------------------------------------
|
72 |
+
# Load the Stable Diffusion pipeline from Hugging Face.
|
73 |
+
# -----------------------------------------------------------------------------
|
74 |
+
MODEL_ID = "stabilityai/stable-diffusion-2" # You may change the model if desired.
|
75 |
+
|
76 |
+
# Use half precision if GPU is available.
|
77 |
+
if torch.cuda.is_available():
|
78 |
+
torch_dtype = torch.float16
|
79 |
+
else:
|
80 |
+
torch_dtype = torch.float32
|
81 |
+
|
82 |
+
print("Loading Stable Diffusion pipeline...")
|
83 |
+
pipe = StableDiffusionPipeline.from_pretrained(MODEL_ID, torch_dtype=torch_dtype)
|
84 |
+
if torch.cuda.is_available():
|
85 |
+
pipe = pipe.to("cuda")
|
86 |
+
print("Stable Diffusion pipeline loaded.")
|
87 |
+
|
88 |
+
# -----------------------------------------------------------------------------
|
89 |
+
# Create the Gradio Interface.
|
90 |
+
# -----------------------------------------------------------------------------
|
91 |
+
title = "Background Removal & Replacement"
|
92 |
+
description = (
|
93 |
+
"Upload an image (e.g., a person or an animal) and provide a text prompt "
|
94 |
+
"describing the new background. The app will remove the original background and "
|
95 |
+
"composite the subject onto a generated background."
|
96 |
+
)
|
97 |
+
|
98 |
+
iface = gr.Interface(
|
99 |
+
fn=process_image,
|
100 |
+
inputs=[
|
101 |
+
gr.inputs.Image(type="pil", label="Upload Your Image"),
|
102 |
+
gr.inputs.Textbox(lines=2, placeholder="Describe the new background...", label="Background Prompt")
|
103 |
+
],
|
104 |
+
outputs=gr.outputs.Image(type="pil", label="Output Image"),
|
105 |
+
title=title,
|
106 |
+
description=description,
|
107 |
+
allow_flagging="never"
|
108 |
+
)
|
109 |
+
|
110 |
+
# -----------------------------------------------------------------------------
|
111 |
+
# Launch the app.
|
112 |
+
# -----------------------------------------------------------------------------
|
113 |
+
if __name__ == "__main__":
|
114 |
+
iface.launch()
|